The CELF family of RNA binding proteins is implicated in cell-specific and developmentally regulated alternative splicing

Abstract

Alternative splicing of cardiac troponin T (cTNT) exon 5 undergoes a developmentally regulated switch such that exon inclusion predominates in embryonic, but not adult, striated muscle. We previously described four muscle-specific splicing enhancers (MSEs) within introns flanking exon 5 in chicken cTNT that are both necessary and sufficient for exon inclusion in embryonic muscle. We also demonstrated that CUG-binding protein (CUG-BP) binds a conserved CUG motif within a human cTNT MSE and positively regulates MSE-dependent exon inclusion. Here we report that CUG-BP is one of a novel family of developmentally regulated RNA binding proteins that includes embryonically lethal abnormal vision-type RNA binding protein 3 (ETR-3). This family, which we call CELF proteins for CUG-BP- and ETR-3-like factors, specifically bound MSE-containing RNAs in vitro and activated MSE-dependent exon inclusion of cTNT minigenes in vivo. The expression of two CELF proteins is highly restricted to brain. CUG-BP, ETR-3, and CELF4 are more broadly expressed, and expression is developmentally regulated in striated muscle and brain. Changes in the level of expression and isoforms of ETR-3 in two different developmental systems correlated with regulated changes in cTNT splicing. A switch from cTNT exon skipping to inclusion tightly correlated with induction of ETR-3 protein expression during differentiation of C2C12 myoblasts. During heart development, the switch in cTNT splicing correlated with a transition in ETR-3 protein isoforms. We propose that ETR-3 is a major regulator of cTNT alternative splicing and that the CELF family plays an important regulatory role in cell-specific alternative splicing during normal development and disease.

FIG. 1

The CELF family members are closely related. (A) All of the CELF proteins possess the same domain structure: three RRMs and a divergent domain of unknown function between RRM2 and RRM3. Nomenclature is designed to consider CUG-BP and ETR-3 as CELF1 and CELF2, respectively. (B) Sequences of five human CELF proteins derived from PCR-amplified cDNAs (see Materials and Methods). Blue, conserved nonpolar amino acids; red, conserved uncharged polar residues; green, conserved positively charged residues; gold, conserved negatively charged residues; black, nonconserved residues. Known alternatively spliced regions are shaded. Italic's, a region of allelic variation in CELF3 due to a variable number of CAG repeats. (C) Phylogenetic relationship of human CELF proteins.

FIG. 1

The CELF family members are closely related. (A) All of the CELF proteins possess the same domain structure: three RRMs and a divergent domain of unknown function between RRM2 and RRM3. Nomenclature is designed to consider CUG-BP and ETR-3 as CELF1 and CELF2, respectively. (B) Sequences of five human CELF proteins derived from PCR-amplified cDNAs (see Materials and Methods). Blue, conserved nonpolar amino acids; red, conserved uncharged polar residues; green, conserved positively charged residues; gold, conserved negatively charged residues; black, nonconserved residues. Known alternatively spliced regions are shaded. Italic's, a region of allelic variation in CELF3 due to a variable number of CAG repeats. (C) Phylogenetic relationship of human CELF proteins.

FIG. 2

CELF proteins are widely expressed. Western blot analysis was performed on total protein samples extracted from adult, pregnant or postpartum female Swiss Webster mice. Fifty micrograms of total protein was loaded in each lane. ETR-3 and CELF4 polyclonal antipeptide antibodies failed to recognize other CELF proteins as transiently expressed or recombinant proteins. An abundant 30-kDa protein was also detected in brain samples on CELF4 blots (data not shown). The relative intensity of this band was reduced somewhat by the addition of a protease inhibitor cocktail to the homogenization buffer, suggesting abundant CELF4 protein expression and cleavage or degradation in brain.

FIG. 3

CELF proteins bind specifically to a downstream MSE-containing region. UV cross-linking experiments were performed using purified recombinant GST-CELF fusion proteins and uniformly ³²P-labeled RNA (G and U) containing partial or full-length segments of the cTNT-regulated region. Proteins and RNA were preincubated in HeLa nuclear extract to increase the specificity of binding. Similar results were obtained when tRNA, bovine serum albumin, and heparin were used instead of HeLa nuclear extract (not shown). Each lane contains equal molar amounts of RNA. An approximately 100-kDa protein that bound nonspecifically to all three RNA substrates in HeLa cell nuclear extracts is also shown.

FIG. 4

CELF proteins promote MSE-dependent exon inclusion. (A) The MSE-containing R35C minigene contains an alternative exon flanked upstream by the last 99 nucleotides of cTNT intron 4, which includes MSE1, and downstream by the first 142 nucleotides of cTNT intron 5, which includes MSEs 2 to 4. This intron-exon-intron cassette is inserted between exons 2 and 4 of the constitutively spliced chicken skeletal troponin I gene. RSV, Rous sarcoma virus. (B) Quail QT35 fibroblasts were cotransfected with 2 μg of R35C and 0 (−), 10, 30, 100, or 300 ng of CUG-BP, ETR-3, CELF5, CELF4, or CELF3 expression plasmid. Increasing concentrations of expression vector showed increasing amounts of protein expression (data not shown). RNA was harvested after 48 h, and RT-PCR analysis was performed to determine the extent of exon inclusion. (C) The MSE-lacking substrate RTBPSRAX contains the same alternative exon as R35C flanked upstream by the last 78 nucleotides of human β-globin intron 1 and downstream by the first 96 nucleotides of the same intron. (D) QT35 cells were cotransfected with 2 μg of RTBPSRAX alone (−) and with 300 ng of CELF5, CELF4, CELF3, ETR-3, or CUG-BP expression plasmid. Asterisk, cryptic splice observed in some PCRs in which the exon is included but the 30-nucleotide exon upstream of the alternative exon is skipped. The appearance of this cryptic splice is inconsistent and represents at most 15% of spliced mRNAs. It is unknown why this cryptic splice is enhanced most efficiently by ETR-3, but enhancement does not appear to be MSE dependent as it is observed for both MSE-containing and MSE-lacking minigenes. Loading was adjusted so that approximately equal counts were loaded in each lane.

FIG. 5

CELF3, CELF4, and to a lesser extent CELF5, but not CUG-BP or ETR-3, can promote exon inclusion via MSE2 alone. (A) The M2/M2TB minigene contains a 52-nucleotide exon flanked upstream and downstream by three copies of the 39-nucleotide cTNT MSE2 from intron 5. RSV, Rous sarcoma virus. (B) QT35 cells were cotranfected with 2 μg of the M2/M2TB minigene and 0 (−), 10, 30, 100, or 300 ng of CELF5, CELF4, CELF3, CUG-BP, or ETR-3 expression plasmid. RT-PCR analysis was performed on RNA harvested 48 h posttransfection. Loading was adjusted so that approximately equal counts were loaded in each lane.

FIG. 6

Developmental profile of CELF proteins in skeletal muscle and brain. Western blot analysis was performed on total protein samples extracted from embryonic-day-14 (e14), newborn (NB), postnatal-day-4 (d4), and adult pregnant or 0- to 4-day-postpartum female Swiss Webster mice. Fifty micrograms of total protein was loaded in each lane. In brain samples, high levels of a 30-kDa protein are observed on CELF4 blots (data not shown), suggesting that a high degree of tissue-specific cleavage or degradation of CELF4 may be occurring in brain as described for Fig. 2.

FIG. 7

An ETR-3 isoform switch that correlates with cTNT splicing is conserved in avian and mammalian heart development. (A) ETR-3 Western blot of total protein samples extracted from embryonic and adult chicken hearts. Sixty micrograms of total proteins was loaded in each lane. (B) Results from primer extension of endogenous cTNT mRNA from the same time course using a ³²P-labeled oligonucleotide complementary to exon 6. (C) ETR-3 Western blot of total protein samples from embryonic-day-14 (e14), newborn (NB), postnatal-day-4 (d4), and adult mouse hearts. Fifty micrograms of total protein was loaded per lane. (D) RT-PCR of endogenous cTNT was performed over the same time course and quantitated by PhosphorImager.

FIG. 8

Induction of ETR-3 protein expression and cTNT exon inclusion occur simultaneously during skeletal muscle differentiation. (A) ETR-3 Western blot of total protein extracts from C2C12 murine myoblast cells induced to differentiate. Fifty micrograms of total protein was loaded in each lane. (B) RT-PCR analysis of cTNT minigene RNA from similar cultures transfected with R35C harvested over the same time course. Loading was adjusted so that approximately equal counts were loaded in each lane.